Effect of different treatment protocols on in vitro fertilisation/intracytoplasmic sperm injection (IVF/ICSI) outcomes in adenomyosis women: a systematic review and meta-analysis

As a benign gynaecological disease, adenomyosis is characterised by the presence of endometrial glands and stroma within the myometrium, which usually results in an enlarged uterus, heavy menstrual bleeding, pelvic pain and infertility. 1 2 Adenomyosis is present in approximately 24.4% of young infertile women in ART(asisted reproductive technolody) cycles. Its prevalence was respectively 38.2% and 34.7% in cases of recurrent pregnancy loss and previous ART failure. 3 Anatomical remodelling, inflammation, immune dysfunction and poor endometrial receptivity have detrimental effects on reproductive outcomes. 4 Moreover, these pathological changes negatively affect fertility, pregnancy and neonatal outcomes in patients with adenomyosis. 5 7 In assisted reproductive technology (ART) cycles, different ovarian stimulation protocols can be adopted. As important confounding factors, protocols have not been systematically evaluated in published systematic reviews and meta-analyses. So far, only the ultralong protocol has attracted attention, and its benefits on pregnancy outcomes have been partly ascertained. 8 11 However, this protocol usually increases the treatment duration and economic cost. Our previously published research demonstrated that the total gonadotropin dosage in the ultralong protocol is 1.28–1.54 times that in other protocols, and the duration of gonadotropin is 2 days longer than that in other protocols. 12 For patients with adenomyosis with poor ovarian reserve, the number of oocytes retrieved using the ultralong protocol may decrease, which may ultimately affect pregnancy outcomes. In addition to the limited knowledge of the ultralong protocol, the advantages and disadvantages of the other protocols are unclear. Based on the above problems, we designed this study and attempted to clarify the following uncertainties: (1) to conduct systematic evaluations of different protocols in in vitro fertilisation/intracytoplasmic sperm injection (IVF/ICSI) cycles; (2) to ascertain a superior transfer strategy, fresh embryo transfer (ET) or frozen ET (FET); (3) to assess the role of GnRH agonist (GnRHa) pretreatment in FET cycles and (4) to provide more flexible treatment strategies.

This systematic review and meta-analysis followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. The registration number is CRD42022340743 (PROSPERO). An electronics-based search was conducted in PubMed, Web of Science, Cochrane library and ClinicalTrials.gov up to October 2023. The following terms were used: “adenomyosis, in vitro fertilization, ART, clinical pregnancy, pregnancy outcomes, miscarriages and live births” ( online supplemental table S1 ). The reference lists of all of the included studies were manually searched. Publications in English or Chinese were also included. All studies were independently reviewed by two reviewers (LG and YL) and discrepancies were resolved by a third investigator (WW). The inclusion criteria were as follows: (1) comparative studies on IVF/ICSI outcomes in an adenomyosis population in peer-reviewed journals, including prospective or retrospective cohort studies and randomised controlled trials (RCTs); (2) diagnosis of adenomyosis with transvaginal ultrasound or MRI according to MUSA (Morphological Uterus Sonographic Assessment) criteria with enlarged and heterogeneous uterus with or without accompanying characteristics, such as asymmetrical thickening, cystic adenomyopathy, hyperechoic islands, fan-shaped shadowing, echogenic subendometrial lines and buds, translesional vascularity, irregular junctional zone and interrupted junctional zone; the diagnosis of non-adenomyosis must exclude these typical ultrasound manifestations of adenomyosis and (3) original publications with unabridged clinical data. The exclusion criteria were as follows: (1) preimplantation genetic testing (PGT), (2) studies mainly concentrating on the effect comparison of surgery and (3) reviews, case reports and animal experiments. The protocols included ultralong, long, antagonist and short. The ultralong protocol involved the administration of long-acting GnRHa (3.75/3.6 mg) on day 2 or 3 of the menstrual cycle. After 28 days, ultrasound was performed and serum hormone levels were examined. When standard downregulation was achieved, gonadotropin was administered until the HCG trigger day. The long protocol involved receiving a daily dose of GnRHa (0.05–0.1 mg) or buserelin nasal spray four times a day at 6-hour intervals (800 µg/day) for pituitary downregulation which started in the mid-luteal phase till the HCG trigger day. After 14 days, gonadotropin was added until the HCG trigger day if standard downregulation was achieved. The antagonist protocol involved the administration of gonadotropin on days 2–4 of the menstrual cycle, before adding a daily dose of GnRH antagonist (0.25 mg) on the fifth or sixth day until the HCG trigger day. The short protocol involved administration of a daily dose of GnRHa (0.05–0.1 mg) on days 2–4 of the menstrual cycle until the HCG trigger day. After 1–2 days of GnRHa treatment, gonadotropin was added for ovarian stimulation, which lasted for approximately 8–12 days. None. Valid information was extracted by two independent authors (LG and YL) according to a standard data format, including authors, year of publication, country, study design, age, protocols in adenomyotic and non-adenomyotic populations, cycles, mode of embryo transfer (ET), diagnostic method and outcome variables. The primary outcome was the clinical pregnancy rate (CPR), and the secondary outcomes were the miscarriage rate (MR) and live birth rate (LBR). Cohort studies were examined by two independent reviewers (WW and ZL) using the Newcastle-Ottawa Scale (NOS). The scoring system included three main parts—selection, comparability and exposure/outcome—with eight items. The quantitative quality appraisal of individual studies ranged from 0 to 9. RCTs were assessed by version 2 of the Cochrane tool for assessing risk of bias in randomised trial, which included five parts–evaluations of randomisation process, deviations from the intended, missing outcomes, measurement of the outcome and selection of reported results. Discrepancies were resolved by a third independent reviewer (LC). The CPR, MR and LBR were expressed as ORs with 95% (CIs. 13 Moreover, I 2 was used as the statistical index to describe the heterogeneity of the studies. If I 2 was ≥50%, a random effect model was adopted for pooled effect and sensitivity analysis was conducted. Conversely, when I 2 <50%, a fixed-effects model was used. 14 15 Subgroup analysis was conducted to analyse the pregnancy outcomes of the different treatment protocols. Funnel plots were produced for asymmetric inspection to assess the possible bias in the publications. 16 All analyses were conducted using Review Manager (RevMan, V.5.3), and significance was set at p<0.05.

A flow diagram of the literature search is shown in figure 1 . The search initially identified 1471 studies via databases (550 studies in PubMed, 820 studies in Web of Science and 101 studies in Cochrane library) and 2 studies by citation searching. After excluding duplicates and unreasonable abstracts, 34 studies were included in the eligibility evaluation. Subsequently, 13 ineligible studies were excluded because of the study population with endometriosis (n=2), receiving PGT (n=2), oocyte donation (n=1), no explicit COS (control ovarian stimulation) protocols (n=4), incomplete RCT (n=1), surgery evaluations (n=1), asymptomatic adenomyosis (n=1) or overlapping samples (n=1). Two studies were deemed eligible based on the citation search.( figure 1 ) We included 23 studies with 4872 cycles in the adenomyosis population and 6289 cycles in the non-adenomyosis population. Among the 23 eligible studies, 14 concentrated on comparisons between adenomyosis and non-adenomyosis populations 1017 29 without publication bias ( online supplemental figure S1 ), and 9 studies only focused on subgroup comparisons of adenomyosis populations. 811 12 30 35 The detailed characteristics are shown in online supplemental table S2 . One study was a RCT, 34 and the other included studies were observational studies, with 5 prospective cohort studies 10 20 24 26 27 and 17 retrospective cohort studies. 811 12 17 19 21 In terms of transplant pattern, 13 studies evaluated fresh ET, 1012 17 22 24 25 29 30 32 6 studies focused on FET 23 27 28 31 33 34 and 4 studies involved fresh ET and FET. 8 11 26 35 Evaluation of long-acting GnRHa pretreatment was performed in six studies with fresh ET cycles, 810 12 30 32 four studies with FET cycles 23 31 33 34 and three studies with fresh ET and FET. 8 11 35 Furthermore, the downregulation durations in fresh ET were as follows: two studies reported durations of ≥3 months, 8 10 two studies reported durations of 2–6 months, 11 30 one study reported a duration of 1–6 months 12 and one study reported a duration of 1–3 months. 32 The downregulation duration in FET cycles was as follows: one study reported a duration of 1–2 months, 23 one study reported a duration of 2 months 31 34 and one study reported a duration of ≥1 month. 33 Furthermore, the CPR was defined as follows: 2 studies with intrauterine or ectopic pregnancy diagnosed by ultrasound or pathological tissue, 17 33 19 studies with an intrauterine gestational sac at 6–8 gestational weeks 810 12 18 and 2 studies with no explicit description. 22 28 The MR was reported in 20 studies and the definitions were as follows: 4 studies with pregnancy loss ≤20 weeks, 18 25 26 34 2 studies with pregnancy loss ≤24 weeks, 21 29 5 studies with pregnancy loss ≤28 weeks, 8 10 12 33 35 6 studies with indirect calculation according to data of clinical pregnancy and live birth, 17 22 23 27 28 32 and 1 study without an explicit description. 11 Two studies with pregnancy loss ≤12 weeks or biochemical pregnancy loss were excluded. 19 20 19 studies reported the LBR. 8 1012 17 18 20 The NOS assessments of the included studies are illustrated in onlinesupplemental tables S3 S4 . The diagnostic evidence of adenomyosis and non-adenomyosis was explicit in all studies. One study was an RCT, 34 and the other studies were retrospective or prospective cohort studies with preset study parameters. 21 studies conducted sufficient intragroup comparability, and 1 study performed propensity score matching. 28 In IVF/ICSI cycles, the CPR was reported in 14 studies, with prevalence of 42.47% (1075/2531) in the adenomyosis group and 55.89% (3515/6289) in the non-adenomyosis group. The intragroup heterogeneity was 31% (I 2 ); therefore, a fixed-effects model was adopted. After pooling the results of 14 studies, we found that adenomyosis accounted for a 26% reduction in CPR (OR: 0.74, 95% CI: 0.66 to 0.82, p<0.00001). The MR was reported in 11 studies, with 27.82% (298/1071) of adenomyosis cases vs 13.9% (454/3266) of non-adenomyosis cases. The OR of the MR ranged from 0.49 (95% CI: 0.10 to 2.45) to 7.50 (95% CI: 1.16 to 48.56) without significant intragroup heterogeneity (I 2 =49%). The results of the fixed effect model indicated a significantly increased risk of miscarriage in the adenomyosis population (OR: 1.90, 95% CI: 1.56 to 2.31, p<0.00001). The LBR was reported in 11 studies, with prevalence of 30.72% (738/2402) in the adenomyosis group and 47.77% (2665/5579) in the non-adenomyosis group. The OR varied from 0.35 (95% CI: 0.08 to 1.60) to 1.20 (95% CI: 0.55 to 2.64), with no obvious intergroup heterogeneity (I 2 =46%). The result of the fixed effect model showed that adenomyosis was responsible for a 35% reduction in LBR (OR: 0.65, 95% CI: 0.58 to 0.73, p<0.00001). The detailed data are shown in figure 2 . The intragroup heterogeneity (I 2 ) for CPR, MR and LBR was 31%, 49% and 46%, respectively, with low heterogeneity. Studies focusing on comparisons between adenomyotic and non-adenomyotic populations were divided into four subgroups according to different protocols ( figure 3 ). The first and second subgroups concentrated on comparisons of the fresh ET cycles. Studies in the adenomyosis group that received long-acting GnRHa pretreatment (ultralong protocol) were categorised into the first subgroup, whereas those in the adenomyosis group that received non-GnRHa pretreatment (long/antagonist/short protocols) were assigned to the second subgroup. The third and fourth subgroups evaluated the pregnancy outcomes in the FET cycles. The third group assessed the effect of long-acting GnRHa pretreatment, and the fourth group assessed the effect of non-GnRHa pretreatment on FET cycles. In the first subgroup, the CPR was 42.82% (653/1525) in the adenomyosis group vs 61.3% (2300/3752) in the non-adenomyosis group, with no intragroup heterogeneity (three studies, OR: 0.72, 95% CI: 0.62 to 0.84, p<0.0001, I 2 =0%). In the second subgroup, the CPR was 34.40% (183/532) in the adenomyosis group vs 58.25% (2516/4319) in the non-adenomyosis group (9 studies, OR: 0.58, 95% CI: 0.47 to 0.71, p<0.00001, I 2 =18%). Although the CPR in the adenomyosis group was lower than that in the non-adenomyosis group in the first and second subgroups, the CPR in the second group was significantly decreased. The MR in the adenomyosis group was significantly increased compared with that in the non-adenomyosis group in the first group (2 studies, 32.08% vs 14.50%, OR: 0.17, 95% CI: 0.11 to 0.23, p<0.00001, I 2 =0%). In the second group, although no statistical difference was achieved, the MR was increased in the adenomyosis group (5 studies, 27.75% vs 13.23%, OR: 0.07, 95% CI: –0.01 to 0.14, p=0.07, I 2 =37%) ( online supplemental figure S2 ). The LBR was 30.82% (470/1525) and 52.13% (1956/3752) in the adenomyosis and non-adenomyosis groups, respectively in the first group with a significant difference (3 studies, OR: 0.59, 95% CI: 0.41 to 0.84, p=0.004, I 2 =74%), and the results of sensitivity analysis supported this conclusion (2 studies, 25.20% vs 34.12%, OR: 0.51, 95% CI: 0.41 to 0.63, p<0.00001, I 2 =0%, online supplemental figure S3 ). Similarly, LBR of adenomyosis group in the second subgroup was very poor (6 studies, 24.07% vs 50.82%, OR: 0.56, 95% CI: 0.42 to 0.75, p=0.0001, I 2 =15%). In the third subgroup, only one study was involved. No significant differences were found in the CPR (51.32% vs 43.48%, p=0.31), LBR (44.74% vs 36.96%, p=0.31) or MR (12.82% vs 12.50%, p=0.97) between the adenomyosis and non-adenomyosis groups. In the fourth subgroup, the adenomyosis group did not receive GnRHa pretreatment. Additionally, we found no statistical differences in CPR (4 studies, 50.25% vs 60.10%, OR: 1.03, 95% CI: 0.79 to 1.34, p=0.82, I 2 =0%), MR (4 studies, 30.5% vs 15.54%, OR: 0.08, 95% CI: –0.03 to 0.20, p=0.15, I 2 =57%) and LBR (4 studies, 34.42% vs 50.25%, OR: 0.81, 95% CI: 0.56 to 1.18, p=0.28, I 2 =42%), consistent with the results of sensitivity analysis of the MR-associated studies ( online supplemental figure S2 ). 10 studies compared IVF/ICSI outcomes using different protocols in the adenomyosis population. In these studies, ultralong or non-ultralong (long and antagonist) protocols were adopted in fresh ET cycles, and long-acting GnRHa pretreatment combined with hormone replacement therapy (HRT) or the sole HRT protocol was adopted in FET cycles ( figure 4 ). Among all studies, including fresh ET and FET cycles, 10 studies reported CPR, 8 reported MR and 7 studies reported LBR. The CPR was 51.03% (844/1654) in the long-acting GnRHa pretreatment group and 38.02% (411/1081) in the non-GnRHa pretreatment group. After pooling the results of 10 studies with a random effect model (I 2 =65%), we found that long-acting GnRHa could significantly improve CPR (OR: 1.44, 95% CI: 1.07 to 1.94, p=0.02, I 2 =65%). Sensitivity analysis supported these results (9 studies, 50.96% vs 40.06%, OR: 1.30, 95% CI: 1.09 to 1.56, p=0.004, I 2 =39 % , online supplemental figure S4 ). The difference in MR between the two groups was not statistically significant (8 studies, 22.14% vs 26.16%, OR: 0.96, 95% CI: 0.70 to 1.32, p=0.82, I 2 =2 % ). Sensitivity analysis indicated that the results were stable. The initial analysis revealed no significant difference in LBR between the groups (seven studies, 41.65% vs 31.79%, OR: 0.04, 95% CI: –0.04–0.13, p=0.31, I 2 =69 % ). However, following sensitivity analysis, we observed that the LBR in the long-acting GnRHa group was higher than that in the non-long GnRHa group (seven studies, 42.54% vs 30.20%, OR: 0.08, 95% CI: 0.03 to 0.12, p=0.002, I 2 =44%, online supplemental figure S5 ). Therefore, long-acting GnRHa therapy may improve CPR and LBR. In fresh ET cycles, compared with long/antagonist protocols, the CPR could be improved by the ultralong protocol (6 studies, 52.89% vs 40.87%, OR: 1.33, 95% CI: 1.06 to 1.66, p=0.01, I 2 =40%). There was no existence of differences in MR (5 studies, 22.00% vs 21.09%, OR: 1.06, 95% CI: 0.71 to 1.58, p=0.77, I 2 =31%) and LBR (4 studies, 44.28% vs 37.18%, OR: 1.08, 95% CI: 0.65 to 1.81, p=0.76, I 2 =72%). Because of the intragroup heterogeneity in LBR, we made sensitivity analysis and the results indicated that the benefit of ultralong protocol on LBR (3 studies, 46.04% vs 36.93%, OR: 1.40, 95% CI: 1.06 to 1.86, p=0.02, I 2 =26%). The results are presented in onlinesupplemental figure S5A S6 undefined . This ultralong protocol may be beneficial for fresh ET cycles. Only four studies focused on FET cycles, and all studies compared GnRHa pretreatment combined with HRT and HRT alone. Four studies reported the CPR, and three reported the MR and LBR. After pooling the results of the studies, we observed no statistical differences in CPR (4 studies, 45.05% vs 33.79%, OR: 1.57, 95% CI: 0.81 to 3.06, p=0.18, I 2 =81%), MR (3 studies, 30.1% vs 37.84%, OR: 0.87, 95% CI: 0.49 to 1.52, p=0.62, I 2 =0%) and LBR (3 studies, 28.57% vs 23.1%, OR: 1.25, 95% CI: 0.86 to 1.82, p=0.25, I 2 =41%). The details are shown in figure 5B . We conducted sensitivity analysis to adjust intragroup differences and no obvious effect was observed (3 studies, 40.86% vs 38.28%, OR: 1.09, 95% CI: 0.78 to 1.53, p=0.60, I 2 =17%, online supplemental figure S7 ). In the FET cycles, CPR and LBR with long-acting GnRHa pretreatment seemed higher than those without GnRHa pretreatment; however, no statistical differences were observed. Which transfer strategy is superior, fresh ET or FET? First, we compared the ultralong protocol in fresh ET cycles and GnRHa pretreatment in FET cycles. No differences were found in CPR (3 studies, 53.94% vs 49.2%, OR: 1.04, 95% CI: 0.61 to 1.76, p=0.89, I 2 =60%), MR (3 studies, 18.54% vs 23.83%, OR: 0.77, 95% CI: 0.47 to 1.26, p=0.29, I 2 =16%) and LBR (2 studies, 46.34% vs 41.82%, p=0.96, I 2 =82%). The results of sensitivity analysis were similar (2 studies, 55.98% vs 46.98%, OR: 1.30, 95% CI: 0.94 to 1.80, p=0.12, I 2 =0%, online supplemental figure S8 ). Subsequently, we compared GnRHa pretreatment in FET cycles with no GnRHa pretreatment (long/antagonist protocol) in fresh ET cycles. GnRHa pretreatment in FET cycles had higher CPR (2 studies, 55.98% vs 33.21%, OR: 1.91, 95% CI: 1.30 to 2.80, p=0.001, I 2 =0%) and LBR (one study, 50.93% vs 33.91%, p=0.003) while no difference in MR was observed (two studies, 15.86% vs 22.99%, OR: 0.79, 95% CI: 0.39 to 1.59, p=0.50, I 2 =55%). The detailed data are shown in figure 5C . GnRHa pretreatment combined with FET appears to be more attractive.

In our study, we systematically evaluated the IVF/ICSI outcomes of different treatment strategies in individuals with adenomyosis. This involved comparisons between adenomyosis and non-adenomyosis groups, as well as among subgroups within the adenomyosis population. Compared with the non-adenomyosis population, adenomyosis might be responsible for a 26% reduction in CPR, 35% reduction in LBR and a 1.9-fold increase in MR. In the analysis between the adenomyosis and non-adenomyosis groups, the ultralong protocol could improve CPR and LBR of the adenomyosis group in fresh ET cycles; however, long/short/antagonist protocols had poor performance with a decreased CPR, LBR and increased MR. Therefore, if women with adenomyosis desire to receive fresh ET, long/short/antagonist protocols should be adopted with caution. Protocol comparisons among adenomyosis in fresh ET cycles confirmed the benefits of ultralong protocols. In FET cycles, GnRHa combined with FET might be a better choice for women with adenomyosis. However, the effect of long-acting GnRHa requires further verification because of the limited number of studies. In fresh ET and FET cycles, the MR in the adenomyosis group was high and was difficult to reverse using long-acting GnRHa. The mechanisms underlying adenomyosis-associated infertility are complex and involve genetic and epigenetic alterations, ovarian steroid hormone aberrations, immune disorders, and inflammatory alterations, with hyperestrogenism and progesterone resistance representing important pathological characteristics. 4 Long-acting GnRHa can induce a hypoestrogenic status and is widely used in women with adenomyosis as a pretreatment protocol in IVF/ICSI cycles. Khan et al reported that GnRHa reduced inflammatory reactions, inhibited angiogenesis and induced apoptosis in ectopic foci. 36 Abnormal endometrial peristalsis and endometrial receptivity can be partly reversed by GnRHa-mediated hypoestrogen status; therefore, improvements in the implantation rate and CPR were achieved in patients with adenomyosis undergoing GnRHa downregulation. 8 1130 32 In this systematic review and meta-analysis, we observed that MR was difficult to reverse using GnRHa pretreatment. The mechanisms of miscarriage are complex and difficult to explain using a single factor. Uterine cavity distortion and abnormal endometrial peristalsis may be linked with these miscarriages. Additionally, the overexpression of the oxytocin receptor (OTR) is an influencing factor. Zhai et al showed that the OTR was expressed in normal endometrial cells and myocytes in the junction zone and outer myometrium, which varied according to the cycle phase and pregnancy. 4 Furthermore, OTR expression positively correlates with the severity of adenomyosis, 37 which leads to implantation disorders and defective deep placentation. 38 With the extension of pregnancy, the biological action of GnRHa gradually vanishes, whereas the OTR in the myometrium increases, which may explain why the MR is difficult to reverse using GnRHa in patients with adenomyosis. An ultralong protocol is usually used in patients with adenomyosis and good ovarian function. In fresh ET cycles, the ultralong protocol seemed to improve pregnancy outcomes; however, increased gonadotropin duration and dosage increased the economic burden. In the FET cycle, long-acting GnRHa appeared to have no effect on pregnancy outcomes. In patients with poor ovarian function, an ultralong protocol should be adopted with caution. Long/short/antagonist protocols in fresh ET cycles combined with a flexible FET strategy might be a viable choice for patients with adenomyosis and poor ovarian reserve. Our study has several strengths. This comprehensive meta-analysis systematically evaluated the effects of different protocols on pregnancy outcomes in patients with adenomyosis. A detailed subgroup analysis of fresh ET and FET was performed to provide evidence for the choice of transfer strategy. We found that GnRHa pretreatment combined with FET may be beneficial, although this conclusion requires further confirmation. However, this study has some limitations. Only publications written in English or Chinese were included. Studies focusing on FET cycles were insufficient, which could have resulted in bias in the pooling results. The included studies were retrospective or prospective cohort studies, and bias in age distribution, extent of adenomyosis and coexisting endometriosis might exist. Studies with explicit statements, such as ‘adenomyosis combined with endometriosis’, ‘endometriosis or adenomyosis’ in titles, methods or other parts were excluded. The included studies were retrospective or prospective cohort studies; thus, bias associated with the age distribution, extent of adenomyosis and coexisting endometriosis might exist. More multicentre RCTs or prospective cohort studies should be conducted to verify our conclusions. Although the ultralong protocol can improve the CPR and LBR, it usually increases the treatment duration and economic cost. GnRHa pretreatment combined with FET might be a better choice for women with adenomyosis; however, this requires further confirmation. In fresh ET cycles, long/short/antagonist protocols should be prudently adopted because of the poor CPR and LBR.